## Reciprocals and Geometry-sensitivity: An experimental study

##### Summary

Reciprocal sentences contain a reciprocal expression, a reciprocal antecedent and a relation (e.g. John, Mary and Sue know each other). The reciprocal expression each other is known to receive a wide variety of interpretations, depending on the predicate in its scope. Previous accounts of reciprocity aim to predict the interpretation of any given reciprocal by selecting a maximal or maximally typical option (e.g. Dalrymple et al., 1998; Sabato, 2006; Kerem, Friedmann & Winter, 2009). In such proposals, maximality is defined in terms of the number of relations among the individuals that make up the antecedent set. This thesis proposes a geometry-sensitive hypothesis, in which the spatial configuration (“geometry”) of those individuals is taken into account as an additional factor.
We provide data for which previous accounts either make incorrect predictions or cannot give a formalized explanation. These data concern reciprocal sentences with two possible interpretations containing unequal amounts of relations and different spatial configurations of individuals (a line and a closed circle). The geometry-sensitive hypothesis that is put forward, determines maximality given a configuration and consequently predicts both of the attested interpretations to be maximal.
Geometry-sensitivity was tested for 22 predicates in an experiment with 71 native speakers of Dutch. We measured acceptability rates of reciprocal sentences in two set-theoretically equivalent situations that differed merely in configuration. The compared situations contained an equal number of individuals (three) and an equal number of relations between them (two), but had the individuals standing either in a line or a circle. Previous accounts that determine maximality only on the basis of number of relations consider the two situations equivalent. For a geometry-sensitive hypothesis, two relations are expected to be maximal given the line configuration but not given the circle configuration.
We found that overall both situations were unacceptable. Geometry-sensitivity was only suggested by results on a subset of the predicates that were tested, namely symmetric predicates. Based on these results, we conclude that geometry-sensitivity does not apply in the broad sense that we initially assumed but possibly in a more restricted way, and we revise our hypothesis accordingly. Finally, we provide suggestions for further research that tests geometry-sensitivity in the revised formulation.